Unit performance health monitoring method and system based on historical data

Abstract

The invention relates to a unit performance health monitoring method and system based on historical data. The method comprises the following steps: collecting and storing operation data, and carrying out preprocessing and feature extraction on the data; a monitoring state evaluation model based on LSTM is adopted, and the health state is evaluated and stored in real time through a doorway mechanism and a memory unit; and according to the obtained health state evaluation structure, analyzing the performance degradation trend and rule of the unit to predict the health state change trend of the unit, and generating personalized maintenance suggestions for the user. According to the invention, multi-source data can be comprehensively collected, advanced data processing and modeling technologies are used, and data-driven performance degradation modeling and intelligent health assessment and maintenance strategy generation are realized.

Description

Technical Field

[0001] The invention relates to a method and system for monitoring unit performance health based on historical data, and belongs to the field of data processing. Background Art

[0002] As the complexity of industrial equipment increases, traditional regular maintenance and manual diagnosis methods can no longer meet the needs of efficient operation of key units. In order to reduce unplanned downtime and maintenance costs and extend equipment life, performance health monitoring technology based on historical data has gradually become one of the key areas. Summary of the invention

[0003] The present invention provides a unit performance health monitoring method and system based on historical data, aiming to solve at least one of the technical problems existing in the prior art.

[0004] The technical solution of the present invention relates to a unit performance health monitoring method based on historical data. The method according to the present invention comprises the following steps:

[0005] S100, collecting and storing operation data, and performing preprocessing and feature extraction on the data;

[0006] S200, adopts LSTM-based monitoring status evaluation model, through the door mechanism and memory unit, to evaluate and store the health status in real time;

[0007] S300: Analyze the performance degradation trend and law of the unit according to the obtained health status evaluation structure, and generate personalized maintenance suggestions for users based on the predicted health status change trend of the unit.

[0008] Furthermore, in step S100, the collected operating data include performance parameters, key data related to the rotor and maintenance data; the performance parameters include vibration, speed, load and temperature; the key data related to the rotor include historical vibration and load change records.

[0009] Furthermore, in step S100, principal component analysis is used to extract key features that affect the health status of the unit from the multi-dimensional sensor data.

[0010] Furthermore, for the step S100, the key features include vibration mode and frequency components, average temperature and temperature fluctuation amplitude, speed fluctuation range and fluctuation frequency, and load average value and fluctuation amplitude.

[0011] Furthermore, in step S200: the gate control mechanism includes a forget gate, an input gate and an output gate.

[0012] Furthermore, the LSTM-based model algorithm is expressed as follows:

[0013] Forget Gate:

[0014] f t =σ(W f ·[h t-1 , x t ]+b f )

[0015] Input Gate:

[0016] i t =o(W i ·[h t-1 , x t ]+b i )

[0017]

[0018] Output Gate:

[0019] o t =σ(W o ·[h t-1 , x t ]+b o )

[0020] The final hidden layer output:

[0021]

[0022] h t =o t ⊙tanh(C t )

[0023] Among them, W f , W i , W c , W o are the weight matrices of the forget gate, input gate, output gate, and long-term state respectively; σ(x) is the sigmoid activation function, b f 、b i 、b c 、b o is the corresponding bias vector; tanh(x) is the tanh activation function; ⊙ represents the memory unit; t represents the time, x t represents the model input at time t, h t-1 represents the hidden state of the input at time t-1, C t-1 represents the memory of the model input at time t-1, f t represents the output value of the forget gate at time t, i t represents the output value of the input gate at time t, represents the candidate memory output value at time t, o t represents the output value of the input gate at time t, Ct represents the model output memory at time t, h t Represents the hidden state of the model output at time t;

[0024] Furthermore, in step S200: the final hidden layer is obtained by the following calculation:

[0025] h (i) =h t ·W (i) +b (i)

[0026] Among them, W i is the weight matrix from the input layer to the hidden layer, with a dimension of n×d i , d i is the number of neurons in the hidden layer, b i is the bias vector of the hidden layer, with dimension 1×d i .

[0027] Furthermore, the output layer uses the Sigmoid activation function to obtain the final output:

[0028]

[0029] Furthermore, the mean squared error (MSE) loss function is used to calculate the predicted value A (i) and the true label Y i The loss between:

[0030]

[0031] In the formula, m is the number of samples, is the predicted output of the i-th sample, Y i is the actual label of the i-th sample.

[0032] Furthermore, the connection weight and the bias are updated by the gradient descent method; the gradient descent method is calculated as follows:

[0033]

[0034] In the formula, α is the learning rate and L is the loss function.

[0035] Furthermore, a degradation curve diagram is drawn according to the obtained change trend of the unit health status.

[0036] Furthermore, the maintenance suggestion includes the time period in which maintenance is required, key components that require attention, and maintenance methods.

[0037] The technical solution of the present invention also relates to a computer-readable storage medium on which program instructions are stored, and the above-mentioned method is implemented when the program instructions are executed by a processor.

[0038] The technical solution of the present invention also relates to a unit performance health monitoring system based on historical data, wherein the system comprises a computer device, and the computer device comprises the above-mentioned computer-readable storage medium.

[0039] The beneficial effects of the present invention are as follows:

[0040] The unit performance health monitoring method and system based on historical data of the present invention can comprehensively collect multi-source data, use advanced data processing and modeling technology, data-driven performance degradation modeling and intelligent health assessment and maintenance strategy generation, improve the safety and stability of equipment operation, and solve problems such as missing data, excessive noise, and analysis bias caused by incomplete data. The present invention uses performance degradation models and time series analysis to clearly depict degradation patterns and identify potential performance degradation risks in advance. Based on the combination of high-precision sensors and machine learning algorithms, the system evaluation accuracy is significantly improved, and the monitoring accuracy is improved. Through accurate predictions and suggestions, excessive maintenance is avoided, and maintenance frequency and material waste are reduced. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a basic flow chart of the unit performance health monitoring method according to the present invention.

[0042] Figure 2 Schematic diagram of the structure of the monitoring state evaluation model based on LSTM according to the method of the present invention.

[0043] Figure 3 is a curve diagram of the vibration amplitude of the unit according to an embodiment of the present invention.

[0044] Figure 4 is a curve diagram of unit temperature according to an embodiment of the present invention.

[0045] Figure 5 is a curve diagram of unit load according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] The concept, specific structure and technical effects of the present invention will be clearly and completely described below in combination with the embodiments and drawings to fully understand the purpose, scheme and effect of the present invention.

[0047] It should be noted that, unless otherwise specified, when a feature is referred to as being "fixed" or "connected" to another feature, it may be directly fixed or connected to another feature, or it may be indirectly fixed or connected to another feature. The singular forms of "," "said" and "the" used herein are also intended to include the plural forms, unless the context clearly indicates otherwise. In addition, unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those skilled in the art. The terms used in this specification are intended to describe specific embodiments only and are not intended to limit the invention. The term "and / or" used herein includes any combination of one or more of the related listed items.

[0048] It should be understood that, although the term first, second, third etc. may be adopted to describe various elements in the present disclosure, these elements should not be limited to these terms. These terms are only used to distinguish the same type of elements from each other. For example, without departing from the scope of the present disclosure, the first element may also be referred to as the second element, and similarly, the second element may also be referred to as the first element. The use of any and all examples or exemplary language ("for example", "such as" etc.) provided herein is only intended to better illustrate embodiments of the present invention, and unless otherwise required, the scope of the present invention will not be limited.

[0049] Reference Figures 1 to 5 In some embodiments, the unit performance health monitoring method based on historical data according to the present invention at least includes the following steps:

[0050] S100, collecting and storing operation data, and performing preprocessing and feature extraction on the data;

[0051] S200, adopts LSTM-based monitoring status evaluation model, through the door mechanism and memory unit, to evaluate and store the health status in real time;

[0052] S300: Analyze the performance degradation trend and law of the unit according to the obtained health status evaluation structure, and generate personalized maintenance suggestions for users based on the predicted health status change trend of the unit.

[0053] The unit performance health monitoring system based on historical data of the present invention includes a data acquisition module, a data preprocessing module, a feature extraction module, a health status assessment module, a performance degradation model establishment module and a maintenance suggestion module, which can be applied to 9E units or other similar units, and improve the safety and stability of equipment operation through historical performance data analysis and health status monitoring system. The present invention can comprehensively collect multi-source data, use advanced data processing and modeling technology, data-driven performance degradation modeling and intelligent health assessment and maintenance strategy generation, and improve the safety and stability of equipment operation. It should be noted that the present invention supports model transplantation to other unit types, such as industrial boilers, wind turbines, etc.

[0054] In some embodiments, the data acquisition module uses high-precision sensors and data transmission interfaces to automatically collect performance parameters such as vibration, speed, load, temperature, and key data related to the rotor, such as historical vibration and load change records, and supports maintenance data entry, where the entered maintenance data includes maintenance time and maintenance content, etc. The data acquisition module stores the collected data in the system database.

[0055] In some embodiments, the data preprocessing module preprocesses the data, uses interpolation to fill in missing values ​​in the time series, maintains the continuity of the time series, and then uses a moving average filter to smooth the time series to reduce noise interference. Then, an outlier detection algorithm is used to detect abnormal points in the time series. The abnormal values, missing values, and duplicate values ​​in the collected data are automatically processed. Among them, the interpolation methods used include existing linear interpolation and spline interpolation.

[0056] In some embodiments, the preprocessed clean data is subjected to feature extraction through a feature extraction module. Specifically, the feature extraction module uses principal component analysis (PCA) to extract key features that affect the health status of the unit from multidimensional sensor data such as vibration data, temperature, and rotation speed, reduce data dimensions, remove redundant information, and improve the interpretability of features. A machine learning algorithm is used to extract features from the data, and the feature values ​​in the database are updated in real time to provide input data for health status assessment. Among them, the extracted key features include the main vibration modes and frequency components, the average temperature and temperature fluctuation amplitude, the fluctuation range and fluctuation frequency of the rotation speed, and the average value and fluctuation amplitude of the load.

[0057] In some embodiments, the health status assessment module assesses the health status in real time through algorithms such as LSTM models and decision trees, and outputs the health status assessment results. Figure 2 ,The present invention uses the LSTM model as the health status assessment algorithm.

[0058] It should be noted that LSTM (Long Short-Term Memory) is a powerful algorithm that can capture long-term dependencies in time series data. Compared with traditional RNN (recurrent neural network), LSTM introduces a gating mechanism, including a forget gate, an input gate, and an output gate, which enables the LSTM network to flexibly process information between different time steps and respond more accurately to complex changes in the extracted features. In addition, LSTM introduces memory units, which solve the gradient vanishing or gradient exploding problems of traditional RNN when processing long sequence data by introducing memory units and gating mechanisms. Among them, the introduction of the gating mechanism not only improves the accuracy of the model, but also enables LSTM to adapt more flexibly when facing multi-axis loads with different amplitudes, frequencies, and phases, thereby evaluating the health status of the unit in real time and predicting possible faults or abnormal conditions.

[0059] Specifically, see Figure 2 In the LSTM-based monitoring state evaluation model, the hidden state set H at time t-1 is t-1 and memory set C t-1 , and the input set X at time t t , the three are used as input data, through the forget gate F i , Input Gate I t , candidate memory After processing, it passes through the output gate O t Output the hidden state set H at time t t and memory set C t Among them, the calculation process involved in the LSTM unit is as follows:

[0060] Forget Gate:

[0061] f t =σ(W f ·[h t-1 , x t ]+b f )

[0062] Input Gate:

[0063] i t =σ(W i ·[h t-1 , x t ]+b i )

[0064]

[0065] Output Gate:

[0066] o t =σ(W o·[h t-1 , x t ]+b o )

[0067] Final output:

[0068]

[0069] h t =o t ⊙tanh(C t )

[0070] Among them, W f , W i , W c , W o are the weight matrices of the forget gate, input gate, output gate, and long-term state respectively; σ(x) is the sigmoid activation function, b f 、b i 、b c 、b o is the corresponding bias vector; tanh(x) is the tanh activation function; ⊙ represents the memory unit; t represents the time, x t represents the model input at time t, h t-1 represents the hidden state of the input at time t-1, C t-1 represents the memory of the model input at time t-1, f t represents the output value of the forget gate at time t, i t represents the output value of the input gate at time t, represents the candidate memory output value at time t, o t represents the output value of the input gate at time t, C t represents the model output memory at time t, h t represents the hidden state of the model output at time t, and its final layer weights and biases are obtained by the following calculation:

[0071] h (i) =h t ·W (i) +b (i)

[0072] Among them, W i is the weight matrix from the input layer to the hidden layer, with a dimension of n×d i , d i is the number of neurons in the hidden layer, b i is the bias vector of the hidden layer, with dimension 1×d i , the output layer uses the Sigmoid activation function to get the final output:

[0073]

[0074] The predicted value A is calculated by the mean squared error (MSE) loss function (i) and the true label Y i The loss between:

[0075]

[0076] In the formula, m is the number of samples, is the predicted output of the i-th sample, Y i is the actual label of the i-th sample.

[0077] The connection weights and the biases of the present invention are updated by the gradient descent method, which is calculated as follows:

[0078]

[0079] In the formula, α is the learning rate and L is the loss function.

[0080] In some embodiments, the performance degradation model building module analyzes the performance degradation trends and rules of the unit through historical data, and generates a data-driven performance degradation model to predict the health status change trend of the unit within a certain period of time in the future. The module can draw a degradation curve chart based on the model output results to help users intuitively observe the degradation process of the unit.

[0081] In some embodiments, the maintenance suggestion module can provide users with personalized maintenance suggestions based on the prediction results of the performance degradation model, including the time period for maintenance, key components that need attention, and maintenance methods. Furthermore, the system automatically generates a report based on the maintenance suggestions so that the staff can refer to it and formulate a specific maintenance plan.

[0082] Here is a specific example to illustrate. Figures 3 to 5 , some historical operating data of a 9E unit include: vibration amplitude: 1.2mm / s, 1.5mm / s, 2.0mm / s (time interval 10 days); temperature: 70℃, 75℃, 80℃; load: 85%, 80%, 75%. Using the method of the present invention, features are extracted through PCA analysis, and the health status assessment model predicts the status after 10 days as follows: vibration amplitude prediction value 2.3mm / s, temperature prediction value 85℃, load prediction value 72%, and combined with the performance degradation curve, the system of the present invention generates maintenance suggestions: it is recommended to inspect the rotor and replace key components within the next 15 days.

[0083] It should be appreciated that the method steps in the embodiments of the present invention can be implemented or implemented by computer hardware, a combination of hardware and software, or by computer instructions stored in a non-transitory computer readable memory. The method can use standard programming techniques. Each program can be implemented in a high-level process or object-oriented programming language to communicate with a computer system. However, if necessary, the program can be implemented in an assembly or machine language. In any case, the language can be a compiled or interpreted language. In addition, the program can be run on a programmed ASIC for this purpose.

[0084] Furthermore, the operations of the processes described herein may be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The processes described herein (or variations and / or combinations thereof) may be performed under the control of one or more computer systems configured with executable instructions, and may be implemented as code (e.g., executable instructions, one or more computer programs, or one or more applications) that is executed collectively on one or more processors, by hardware, or a combination thereof. The computer program includes a plurality of instructions that may be executed by one or more processors.

[0085] Further, the method can be implemented in any type of computing platform that is operably connected to a suitable computer, including but not limited to a personal computer, a minicomputer, a mainframe, a workstation, a network or distributed computing environment, a separate or integrated computer platform, or in communication with a charged particle tool or other imaging device, etc. Various aspects of the present invention can be implemented in machine-readable code stored on a non-transitory storage medium or device, whether removable or integrated into a computing platform, such as a hard disk, an optical read and / or write storage medium, an RSM, a ROM, etc., so that it can be read by a programmable computer, and when the storage medium or device is read by the computer, it can be used to configure and operate the computer to perform the process described herein. In addition, the machine-readable code, or portions thereof, can be transmitted via a wired or wireless network. When such media includes instructions or programs that implement the steps described above in conjunction with a microprocessor or other data processor, the invention described herein includes these and other different types of non-transitory computer-readable storage media. When programmed according to the methods and techniques of the present invention, the present invention can also include the computer itself.

[0086] The computer program can be applied to input data to perform the functions described herein, thereby converting the input data to generate output data stored in a non-volatile memory. The output information can also be applied to one or more output devices such as a display. In a preferred embodiment of the present invention, the converted data represents physical and tangible objects, including specific visual depictions of physical and tangible objects produced on the display.

[0087] The above is only a preferred embodiment of the present invention. The present invention is not limited to the above implementation. As long as the technical effect of the present invention is achieved by the same means, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the present invention. Within the scope of protection of the present invention, its technical scheme and / or implementation method may have various modifications and changes.

Claims

1. A unit performance health monitoring method based on historical data, characterized in that: The method comprises the following steps: S100, collecting and storing operation data, and performing preprocessing and feature extraction on the data; S200, adopts LSTM-based monitoring status evaluation model, through the door mechanism and memory unit, to evaluate and store the health status in real time; S300: Analyze the performance degradation trend and law of the unit according to the obtained health status evaluation structure, and generate personalized maintenance suggestions for users based on the predicted health status change trend of the unit.

2. The method according to claim 1, characterized in that In the step S100, The collected operating data include performance parameters, key data related to the rotor and maintenance data; the performance parameters include vibration, speed, load and temperature; the key data related to the rotor include historical vibration and load change records.

3. The method according to claim 1, characterized in that In the step S100, Principal component analysis is used to extract key features that affect the health status of the unit from multidimensional sensor data.

4. The method according to claim 3, characterized in that For step S100, The key features include vibration mode and frequency components, average temperature and temperature fluctuation amplitude, speed fluctuation range and fluctuation frequency, and load average value and fluctuation amplitude.

5. The method according to claim 1, characterized in that In step S200: The gate control mechanism includes a forget gate, an input gate and an output gate.

6. The method according to claim 4, characterized in that n is the number of features of each sample, such as vibration, speed, load, temperature, historical vibration, load change record, etc. The group performance health monitoring method based on the LSTM model algorithm is expressed as follows: Forget Gate: f t =σ(W f ·[h t-1 ,x t ]+b f ) Input Gate: i t =σ(W i ·[h t-1 ,x t ]+b i ) Output Gate: the t =σ(W o ·[h t-1 ,x t ]+b o ) The final hidden layer output is: h t =o t ☉tanh(C t ) Among them, W f , W i , W c , W o are the weight matrices of the forget gate, input gate, output gate, and long-term state respectively; σ(x) is the sigmoid activation function, b f 、b i 、b c 、b o is the corresponding bias vector; tanh(x) is the tanh activation function; ⊙ represents the memory unit; t represents the time, x t represents the model input at time t, h t-1 represents the hidden state of the input at time t-1, C t-1 represents the memory of the model input at time t-1, f t represents the output value of the forget gate at time t, i t represents the output value of the input gate at time t, represents the candidate memory output value at time t, o t represents the output value of the input gate at time t, C t represents the model output memory at time t, h t Represents the hidden state of the model output at time t; The final hidden layer is obtained by the following calculation: h (i) =h t ·W (i) +b (i) Among them, W i is the weight matrix from the input layer to the hidden layer, with a dimension of n×d i , d i is the number of neurons in the hidden layer, b i is the bias vector of the hidden layer, with dimension 1×d i . The output layer uses the Sigmoid activation function to get the final output: The predicted value A is calculated by the mean square error loss function (i) and the true label Y i The loss between: In the formula, m is the number of samples, is the predicted output of the i-th sample, Y i is the actual label of the i-th sample; The output value of the unit performance health monitoring method is obtained according to the connection weight and the bias; the connection weight and the bias are updated by the gradient descent method; the gradient descent method is calculated as follows: In the formula, α is the learning rate and L is the loss function.

7. The method according to claim 1, characterized in that In the step S300, Based on the obtained trend of unit health status changes, a degradation curve is drawn.

8. The method according to claim 1, characterized in that: In the step S400, The maintenance suggestion includes the time period when maintenance is required, key components that require attention, and maintenance methods.

9. A computer-readable storage medium having program instructions stored thereon, wherein the program instructions are executed by a processor to implement the method according to any one of claims 1 to 8.

10. The unit performance health monitoring system based on historical data is characterized by: include: A computer device comprising a computer readable storage medium according to claim 9.

Images